Life Science Research and Data Management – what can they give each other ?

“Databases for the life sciences” is not really a newly emerging area in databases. The Nucleotide Sequence Database from the European Molecular Biology Laboratory (EMBL) has been operational at the European Bioinformatics Institute (EBI) since 1980. SWISS-PROT, the classical database containing protein information, was established in 1986. Today, there are over a thousand different life science databases with contents ranging from gene-property data for different organisms to brain image data for patients with neurological disorders. What then calls for a special issue of the SIGMOD Record devoted to the topic? We may posit three observations that underscore the need for the database community to take a closer and deeper look at the problems associated with life science databases. The first, reported in more detail by Jagadish and Olken in this issue, is that biological information is inherently more complex because it needs to represent objects, interactions and phenomena of the natural world – and yet it has received little attention from the data modeling community ([2] is an interesting exception). Apart from sequence data, there have been few data models to develop efficient representations and operations for biologically relevant data. The second observation is that scientific questions in life sciences lead to much more complex queries than the traditional database world is used to. While significant progress has been made in sequence databases, temporal databases, graph databases, recursive query processing, user-defined aggregate computation and so forth, real-world scientific queries often need to cut across all of them, and place new demands on query languages and evaluation. The third observation follows from the problem of information integration across a combination of homegrown, improperly modeled as well as highly sophisticated databases, often having distinctly different data models, laden with unstated assumptions. In many life science applications, information needs to be integrated across databases that are not only heterogeneous, but may have non-overlapping schemas (because the data are from different instruments or scales of resolution), may have contradictory information (because experiments were performed with different assumptions), or may contain data that can only be compared after matching their context expressed in the form of constraints. Fortunately, the database, data mining and knowledge management communities are gradually becoming more active in investigating these and other life-science inspired problems. Although genomic and proteomic applications dominate the life science data management research today, this is likely to broaden over the next few years and embrace problem areas like disease-specific databases, tissue-specific databases and process-specific databases. In the current issue of the SIGMOD Record, we present some of the research originating from both database and life science application scientists.